Culture-Independent MetagenomicSurveillance of Commercially AvailableProbiotics with High-Throughput Next-Generation Sequencing

Jennifer N Patro Padmini Ramachandran Tammy Barnaba Mark K MammelJada L Lewis Christopher A ElkinsDivision of Molecular Biology Center for Food Safety and Applied Nutrition US Food and DrugAdministration Laurel Maryland USA

ABSTRACT Millions of people consume dietary supplements either following adoctorrsquos recommendation or at their own discretion to improve their overallhealth and well-being This is a rapidly growing trend with an associated andexpanding manufacturing industry to meet the demand for new health-relatedproducts In this study we examined the contents and microbial viability of sev-eral popular probiotic products on the United States market Culture-inde-pendent methods are proving ideal for fast and efficient analysis of foodbornepathogens and their associated microbial communities but may also be relevantfor analyzing probiotics containing mixed microbial constituents These productswere subjected to next-generation whole-genome sequencing and analyzed by acustom in-house-developed k-mer counting method to validate manufacturer la-bel information In addition the batch variability of respective products was ex-amined to determine if any changes in their formulations andor the manufac-turing process occurred Overall the products we tested adhered to theingredient claims and lot-to-lot differences were minimal However there were afew discrepancies in the naming of closely related Lactobacillus and Bifidobacte-rium species whereas one product contained an apparent Enterococcus contami-nant in two of its three lots With the microbial contents of the products identi-fied we used traditional PCR and colony counting methods to comparativelyassess our results and verify the viability of the microbes in these products withregard to the labeling claims Of all the supplements examined only one wasfound to be inaccurate in viability Our use of next-generation sequencing as ananalytical tool clearly demonstrated its utility for quickly analyzing commerciallyavailable products containing multiple microbes to ensure consumer safety

IMPORTANCE The rapidly growing supplement industry operates without a formalpremarket approval process Consumers rely on product labels to be accurate andtrue Those products containing live microbials report both identity and viability onmost product labels This study used next-generation sequencing technology as ananalytical tool in conjunction with classic culture methods to examine the validity ofthe labels on supplement products containing live microbials found in the UnitedStates marketplace Our results show the importance of testing these products foridentity viability and potential contaminants as well as introduce a new culture-independent diagnostic approach for testing these products

KEYWORDS Probiotics dietary supplements metagenomics whole-genomesequencing lactic acid bacteria

Received 29 February 2016 Accepted 9March 2016 Published 30 March 2016

Citation Patro JN Ramachandran P Barnaba TMammel MK Lewis JL Elkins CA 2016 Culture-independent metagenomic surveillance ofcommercially available probiotics with high-throughput next-generation sequencingmSphere 1(2)e00057-16doi101128mSphere00057-16

Editor Garret Suen University of WisconsinMadison

Copyright copy 2016 Patro et al This is an open-access article distributed under the terms ofthe Creative Commons Attribution 40International license

Address correspondence to Christopher AElkins chriselkinsfdahhsgov

JNP and PR contributed equally to this work

Sequencing of probiotics sold in USmarket

RESEARCH ARTICLEApplied and Environmental Science

crossmark

Volume 1 Issue 2 e00057-16 msphereasmorg 1

on May 31 2020 by guest

httpmsphereasm

orgD

ownloaded from

Probiotics are defined by the World Health Organization (WHO) as ldquolive microorgan-isms which when administered in adequate amounts confer a health benefit on the

hostrdquo (1 2) Additionally it has been recommended recently by the InternationalAssociation of Probiotics and Prebiotics that the term ldquoprobioticrdquo be used only forproducts that ldquodeliver live microorganisms with a suitable viable count of well-definedstrains with a reasonable expectation of delivering benefits for the well-being of thehostrdquo (3 4) Microbes can occur naturally in foods be intentionally added or bepackaged solely as a dietary supplement (5) The availability of these products to theconsumer is reflected by the expansion of the dietary supplement business in theUnited States which is now a $30 billion dollaryear global industry (6)

The majority of probiotic products contain Lactobacillus Bifidobacterium Pediococ-cus Streptococcus and Lactococcus species as either a mixture or single isolates Whilethese products report the number of CFU present per serving as well as the type ofmicrobes present in their product on the label they are not required to include thisinformation These types of products are packaged in a variety of ways such as incapsule gummy or powder form Probiotics commonly found on store shelves can bemarketed to specific consumers for example specifically targeting women or childrenThe awareness that probiotics have the potential to affect an individualrsquos health hasincreased over the past several years They function in modulating the host immuneresponse preventing the colonization of pathogenic organisms andor establishing aproper balance of the host gut microbiota (7ndash9) Recently probiotic use has beenextended beyond diet supplementation and has been used in clinical situations (7 10)This alternative application as a medical option raises concerns about the safety ofthese products for human consumption and the possibility that they may requireadditional manufacturing process control Therefore an accurate description of themicrobes present in these products as well as determination of the presence of anymicrobial contaminants specifically those that may pose a health risk is important forpublic safety and the proper labeling of these products

Errors in labeling accuracy due to the miscalculation of levels of bacteria or theproper taxonomic identification of bacterial species have been reported extensivelyworldwide (11ndash17) However most of these studies used PCR or selective culturetechniques which are a priori limited to targeting certain bacteria in a mixed productMoreover such methods are labor intensive and may take many days to complete Onthe contrary species-agnostic approaches better capable of capturing unknown po-tential contaminants or mislabeled components would be optimal especially if con-tained in a single platform Genomic-scale taxonomy has become a popular identifi-cation technique with the cost and speed of whole-genome sequencing rapidlydecreasing over the past several years (18 19) This technique can be used to easilyidentify species present in dietary supplements and determine any mislabeling of theseproducts rapidly and accurately (20) In addition whole-genome sequencing analysis ofprobiotics can determine the presence of any microbial contaminant that can haveadverse effects on public health

In this study we used a culture-independent method involving shotgun next-generation sequencing to determine the contents of various types of commerciallyavailable probiotics This approach will provide an analytical laboratory the ability todetermine in a single assay the types and relative amounts of bacteria present as wellas the presence of bacteria not reported on the label The existence of a microbialcontaminant found during sequencing can be confirmed by PCR with strain-specificprimers andor selective media In addition to the identities of all microbes the viabilityof these probiotic organisms is also of concern Labeling claims commonly include theamount of live microorganisms per capsule Since sequencing cannot determinewhether an organism that is present in a product is viable the accuracy of these claimswas verified by using standard culturing techniques and specific media to isolate anddetermine the viability of all of the microbes in each product

Patro et al

Volume 1 Issue 2 e00057-16 msphereasmorg 2

on May 31 2020 by guest

httpmsphereasm

orgD

ownloaded from

RESULTSSequencing quantitation and limit of detection In order to determine whether theDNA extraction technique used would yield the expected quantitative results twodifferent mixed mock samples containing various amounts of representative specieswere developed (Table 1) In the first method single-colony isolates of known strainswere grown in their respective media overnight and combined in known relativeamounts DNA was extracted from this mixture in the same manner as DNA wasextracted from the probiotic products In the second method DNA was isolated fromeach single-colony isolate and then combined into one sample of known relativeamounts With both methods we comparatively assessed sequencing read quantitationagainst these artificially constructed metagenomic samples

The relative amount of DNA obtained by each approach was determined by k-meranalysis The mock culture mixture was not as accurate as the mock DNA mixture whichwas extremely accurate for strains with abundances of 3 (Table 1) Below this levelthe relative amount of each strain in both constructions was identified with lessaccuracy While the mock culture mixture is less accurate quantitatively the dominantspecies were properly identified and all species present at 3 were detected Thussequencing has the qualitative ability to determine the dominant species present in thesample although quantitation may be dependent on additional undefined factors suchas relative lysis efficiency or inherent DNA stabilities in mixed culture lysis We extendedthese findings with arguably the more complex products used in this study one ofwhich (product G) was extensively characterized for other purposes Thus preliminaryk-mer sequence analysis of products G and J revealed a sharp increase in speciesidentifications below approximately 15 relative sequence abundance (Fig 1) For thepurposes of our study results below this level were considered noise and were scoredappropriately (as not present in the sample) In addition this was concluded becauseof the lack of false positives produced by organisms present at 15 throughout therest of the study

It is important in product surveillance to capture all of the organisms present in thesamples without introducing false positives Therefore the cutoff of species present ina sample was established as 15 with the caveat that quantitation of the relativeabundances of the DNA species present cannot be established below 3 We alsoexamined the limit of detection of a strain particularly relevant to this study in a mixedsample This was achieved by spiking 50 109 CFU of probiotic product G with aknown number of Enterococcus faecium cells (see Table S4 in the supplemental mate-rial) The relative abundance of E faecium was determined by the in-house k-mer

TABLE 1 Percent species and DNA abundances determined by k-mer analysis of mixedmock sample sequences

Species

Abundance

Read Mock

Mixed mock DNALactobacillus acidophilus 416 45Lactobacillus plantarum 355 35Lactobacillus zeae 71 7Lactobacillus helveticus 101 6Lactobacillus brevis 41 3Lactobacillus casei group 00 3Lactobacillus reuteri 15 1

Mixed mock cultureLactobacillus acidophilus 651 45Lactobacillus plantarum 158 35Lactobacillus zeae 105 7Lactobacillus helveticus 41 6Lactobacillus brevis 00 3Lactobacillus casei group 41 3Lactobacillus reuteri 00 1

Metagenomic Sequencing of Probiotics

Volume 1 Issue 2 e00057-16 msphereasmorg 3

on May 31 2020 by guest

httpmsphereasm

orgD

ownloaded from

method described and the results show that the contaminant is detectable when it isused to spike a sample at levels as low as 15 With the exception of the lowestconcentration of E faecium the relative amount was determined to be higher than theactual amount used to spike the product However the amount of E faecium didincrease at increasingly higher spiking levels In addition for other organisms in theproduct relative sequencing yields tended to be consistent (demonstrating precision)across multiple repetitive samples

Metagenomic sequencing of probiotic products Ten of the top-selling onlineprobiotic products were tested for content with Illumina MiSeq These products wereassigned letters A to J The reported ingredients of each product are listed in Table 2The majority of these products have more than one microbe listed as ingredients ontheir labels and identify the microbes present in their product to the species levelTherefore the ability to identify multiple types of bacteria without culturing makes thisan ideal technique for identifying these products in comparison to the product labelingHowever this method could be used to identify supplement contents down to thestrain level if this information is supplied by the manufacturer We present a simplifiedversion of the results in Fig 2 showing bacteria (i) present on the label and detectedvia sequencing (green) (ii) present on the label but not detected via sequencing (red)and (iii) not on the label but detected during sequencing (yellow) In addition somebacteria were also indicated that are below the limit of detection (detected at a relativeabundance of 01 to 15) but could be present in the sample The relative abundancecutoff for a present call is 15 that for a marginal call (represented by a line) is 01 to15 and that for an absent call is 01 For two example products (G and H) therelative amounts of the microbes present are stratified and displayed across differentproduct lots (Fig 3A) as was done for all of the products (see Table S5 in thesupplemental material) Of the 10 products examined 1 (product G) potentially con-tained additional bacteria that were not indicated on the label including a possibleE faecium contaminant (product G lot 1 Fig 2) This finding was further confirmed byculture methods and whole-genome sequencing (see below) In parallel many constit-uent bacteria were assessed by PCR assay and viability testing Thus they weredetectable with species-specific PCR primers (see Table S3 in the supplemental mate-rial) and were also grown on appropriate solid media to confirm their presence (seeTables S6 and S7 in the supplemental material)

Sequencing analysis of these products determined several mislabeling issues al-though many of these were simply apparent errors in nomenclature ie Lactobacillusacidophilus as L helveticus and L casei as L zeae (21ndash23) In addition the label of oneproduct (J) misidentified a common Bifidobacterium species as Bifidobacterium lactis(infantis) instead of B longum subsp infantis (Fig 2) and revealed how frequent andobvious the mislabeling of these products has become Our results show that this is a

FIGURE 1 Bacterial species identified at lt2 relative abundance in products G and J by k-meranalysis

Patro et al

Volume 1 Issue 2 e00057-16 msphereasmorg 4

on May 31 2020 by guest

httpmsphereasm

orgD

ownloaded from

common practice however these organisms have not raised safety concerns in thepast

Enumeration In addition to ingredient labeling most probiotic products containactive-culture information To test claim validity each product was directly plated ontoselective media for each organism identified during sequencing The types of mediaused to examine the presence of each bacterium (see Tables S6 and S7 in thesupplemental material) not only allowed us to determine the number of CFU in eachproduct but also confirmed our sequencing results All of the products contained theamounts of live cultures reported on their labels within 90 except one product H(Table 2) which was reported to contain 100 109 CFU but contained only 80

109 CFUConfirmation of sequencing results The sequencing results revealed some

labeling discrepancies that were further explored by other techniques Two of theproducts (G and H) were examined with our custom-designed Affymetrix microarrayGutProbe (Fig 3B) From our sequencing analysis product H was qualitatively consis-tent across product lots with only limited apparent quantitative differences (Fig 3A)whereas product G had different product formulations in lots 1 and 2 (Table 2footnote c) These observations were readily apparent with the GutProbe platformThus pairwise scatterplot comparisons of product lot formulations that are similarshould display equal hybridization intensities across the microarray resulting in probesets clustering tightly along the diagonal as shown for product H (lots 2 and 3)However formulation differences such as those in product G display biases in hybrid-ization intensities specific to a respective lot (Fig 3B)

In addition all products were plated onto species-specific media in order to deter-mine the morphological presence of bacteria detected via sequencing All species in

TABLE 2 Comparison of product labeling values for viability and identity versusbacteriological culturing conducted in this study

Product Ingredient(s)a CFU countb

A Bifantis contains 1 109 CFU of B infantis 35624 (40 mg)when manufactured and provides an effective level of bacteria(1 107 CFU) until at least the ldquobest-byrdquo date

1 109

B 10 109 cells of Lactobacillus sp strain GG10 109 cells of inulin (chicory root extract) (200 mg)

10 109

C 5 109 cells of Lactobacillus sp strain GG 5 109

D 4 109 cells of L acidophilus B bifidum B longum andL plantarum OM

4 109

E 12 109 CFU of L acidophilus la-14 12 109 CFU ofB lactis BI-04 1 109 CFU of L casei Lc-11 1 109 CFUof B breve Bb-03 1 109 CFU of L salivarius Ls-33)1 109 CFU of L plantarum Lp-115 1 109 CFU ofB longum Bl-05 1 109 CFU of L rhamnosus Lr-32

30 109

F 34 109 CFU of B longum BB536 L acidophilus la-14B lactis BI-04 L rhamnosus R0011 L casei R0215 andL plantarum R1012

3 109

Gc Proprietary blend of 50 109 cells of L paracasei B lactisL plantarum L acidophilus B longum L rhamnosusS thermophilus L casei L salivarius and L reuteri strains

50 109

H 100 109 CFU of B breve 129 B longum 135B bifidum 132 L acidophilus 122 and L rhamnosus 111

80 109

I 25 109 CFU of L rhamnosus GR-1 25 109 CFU ofL reuteri RC-14

5 109

J 32 109 cells of B bifidum B breve B lactis (infantis)B lactis HN019 B longum L acidophilus L brevisL bulgaricus L casei L gasseri L paracasei L plantarumL rhamnosus L salivarius L lactis and S thermophilus

30 109

aAs reportedlabeled on productbExperimentally determined in this studycProduct G contained a change in the formulation on the label in the lots tested in this study Lot 1 isdisplayed here In lots 2 and 3 L reuteri was replaced with B breve

Metagenomic Sequencing of Probiotics

Volume 1 Issue 2 e00057-16 msphereasmorg 5

on May 31 2020 by guest

httpmsphereasm

orgD

ownloaded from

general were distinguished from one another on the basis of colony morphology withrecorded colony descriptions (see Table S6 in the supplemental material) However thecolony morphologies of L salivarius L fermentum L casei and L paracasei were notdiscernible in relation to one another Therefore we were unable to identify those

FIGURE 2 Summary qualitative analysis of metagenomic sequencing of various probiotic products and associated inconsistences with labeling

Patro et al

Volume 1 Issue 2 e00057-16 msphereasmorg 6

on May 31 2020 by guest

httpmsphereasm

orgD

ownloaded from

organisms in two of the products (product G and J) on the basis of colony morphologyCustom-designed primers were used to confirm the diversity of the single-colonyisolates (see Table S3 in the supplemental material) These additional methods are veryimportant in confirming the presence of a contaminant like the one found in lot 1 ofproduct G It was determined via metagenomic sequencing that this sample containeda large relative amount of E faecium that was not detected in the other two lots Asexpected we obtained a single-colony isolate from lot 1 on Enterococcus medium thatwas subsequently sequenced We confirmed that the contaminant was E faecium with80 of the reads mapping to the reference strain of E faecium Unlike the metagenomicresults which only identified E faecium in lot 1 culturing of the two additional lotsrevealed the growth of E faecium in lot 2 Indeed a second isolate was obtained fromlot 2 of product G and the genome sequences of the isolates from both lots wereidentical to one another However they did not match available reference strains inGenBank We further assessed the sequences for virulence factors (24) against the

FIGURE 3 Batch variation of formulations from two products by two different genomic-scale techniques (A)The relative abundance of each bacterium in a product as determined by k-mer analysis of metagenomicsequencing The results are for two examples products G and H across three respective lots (B) Analysis of thesame samples with the custom-designed high-density GutProbe microarray platform (9) Each spot is asummarized probe set intensity (red shading lt85 absent blue shading gt85 present) for a specific genearray from available whole genomes of human gut-associated species including common probiotics As hasbeen shown previously this genomic-scale genotyping is exquisitely sensitive for the determination of theidentities and formulations of products especially when they are assessed in pairwise comparisons

Metagenomic Sequencing of Probiotics

Volume 1 Issue 2 e00057-16 msphereasmorg 7

on May 31 2020 by guest

httpmsphereasm

orgD

ownloaded from

virulence gene database MvirDB (httpmvirdbllnlgov) but failed to identify anymatches

DISCUSSION

The WHO defines probiotics as ldquolive microorganisms which when administered inadequate amounts confer a health benefit on the hostrdquo In todayrsquos market many foodproducts are available with the intentional addition of live microbes with advertisedgeneralized health benefits Ten different probiotic supplements marketed in theUnited States were examined to verify the information presented on the label Amethod to rapidly verify the labeling of these products as well as identify anycontaminants would be useful for quality control and safety Here next-generationsequencing was used to meet these needs Using this method would allow the testingof multiple products in one assay (up to 10 depending on the number of strainsidentified in each product) with results within 1 week While this method is still costlythe expense associated with whole-genome sequencing technologies has been signif-icantly declining over the past decade and continues to fall with the advent ofnext-generation platforms (18) This methodrsquos comprehensive nature makes it worththe initial investment as long as relative ease of postrun analysis continues to improveIn addition since the labeling information on these products usually does not containstrain information or quantitative descriptions of each microbial constituent thismethod could be useful in obtaining additional information beyond a given productlabel

Most of these products were in capsule form and DNA was directly isolated from thepowder contained in the capsules however two of the products were in powder formRegardless since DNA isolation from all products started with a powder form only oneDNA isolation technique was used in this study To establish this method as a viableway to determine relative amounts of bacteria we used a mock sample with knownrelative amounts of bacteria The results reveal that while each component of thesample is identified the relative amounts vary from the known amounts especiallywhen the bacteria are present at 3 On the basis of the mock sample analyses acutoff of 15 was established for the relative amount of bacteria in a sample While theresults are not accurate at relative abundances of 3 the cutoff was chosen becauseof the lack of false positives present at levels of 15 (Table 1 Fig 1) In addition thelimit of detection was further established by spiking a product sample with a knownquantity of E faecium This organism is detected at levels as low as 15 When dietarysupplements were plated to obtain the total colony counts we isolated and identifiedsingle colonies on the basis of their morphologies and confirmed their identities by PCRassay The single-colony isolates were obtained only for organisms where the sequenc-ing reads were 15 This additional effort to confirm identities to the species levelcomplemented the sequencing data to avoid any false-positive calls We realize thatthese conclusions are based on a small data set and they are intended as a first proofof principle to assess the potential of this technique for the identification of probioticbacteria to the species level We also realize that relative lysis methods and efficienciesneed to be further determined given the recalcitrant nature of certain Gram-positiveorganisms

Next-generation DNA sequencing has rapidly become a useful method to identifythe contents of mixed samples (9 25ndash28) and is quite applicable to the identificationof all of the live-microbe ingredients in dietary supplement samples in this study Thevolume of data generated by whole-genome sequencing was analyzed through acustom in-house pipeline Our k-mer method was used to rapidly and effectivelyidentify all of the ingredients in various dietary supplement samples to the specieslevel This approach can be tailored to the desired taxonomic resolution requiredincluding the identity of an organism to the strain level However since most productlabels do not contain this information our results are reported to the species level tocorrespond to the product labeling information Furthermore this analysis method isable to determine the presence of specific contaminants in these products Along these

Patro et al

Volume 1 Issue 2 e00057-16 msphereasmorg 8

on May 31 2020 by guest

httpmsphereasm

orgD

ownloaded from

lines the genomes included in the database are those of bacteria commonly found inthe gut or typically found in marketed products Thus the database does not includecommon food contaminants but can be easily modified to include a wider variety ofbacteria to test products more thoroughly for contamination

Overall high-throughput sequencing data revealed that half of the products ana-lyzed matched the contents listed on their respective labels Of note these samples(products A to D and H) contained the least bacteria all had five or fewer bacterialgenera listed Product A lists B longum subsp infantis on its label and sequencingresults revealed B longum subsp longum These two B longum subspecies are difficultto differentiate taxonomically and this can easily lead to misidentification (29) Inaddition primers were designed to distinguish between these two subspecies byPCR-based methods However product A was positive for both subspecies (infantis andlongum) via PCR indicating that whole-genome sequencing is the best method for theidentification of this productrsquos contents Furthermore sequence analyses highlightedother common nomenclature errors (ie L acidophilus and L helveticus L casei andL zeae) found in multiple products (E H and J) (9 30) However these mislabelingissues are less of a concern than others found in the products tested B bifidum was notdetected in any of the three lots of product J tested (Fig 2) Finally it was apparent thatmanufacturers change their formulations from lot to lot as shown in product G (Fig 3Aand B)

While this method is extremely useful as a screening tool to determine if anymicrobial contaminants are present in products it should not be used as the solemethod to screen live microbial products for contaminants For example sequencingshowed that lot 1 of product G contained an E faecium contaminant that was notdetected in either lot 2 or 3 To further explore this result and confirm the presence ofthis organism classical culture techniques were used with all three lots When appliedto E faecium-specific media both lots 1 and 2 produced growth The growth obtainedwith lot 1 one was 3-fold greater than that obtained with lot 2 (data not shown)potentially revealing the reason why E faecium was not detected by sequencing in lot2 However it is important to note that the sequencing approach did lead to furtherinvestigation of all of the lots of the product to confirm the presence of this bacterium

The numbers of live cultures reported on product labels were tested by traditionalculture-based techniques to verify labeling claims While DNA sequencing can deter-mine whether signatures of a specific organism are present it may not determineviability by our approach however RNA profiling may have future utility in this regardTherefore it is necessary to use basic culture techniques as a secondary verification toolfor sequencing approaches All of the labels tested except for that of product H wereaccurate This method is still extremely labor intensive and requires multiple media tocover all of the bacteria present in each product Each product was stored at 4degC for nomore than 8 months prior to testing It would be of interest to determine the shelf livesof these products at various storage temperatures

The methods currently used to identify microbes within dietary supplements arelaborious and do not provide immediate identification to the species level whendealing with a mixed microbial community Whole-genome sequencing offers anefficient culture-independent method for quickly analyzing a complex sample It canaccurately identify all of the labeled ingredients inclusive of any potential pathogensor contaminants given appropriate depth and breadth of the associated bioinformaticdatabase However at this point it is not unilaterally capable of accurately identifyingall of the microbes in a sample In addition whole-genome sequencing cannot yetestablish the viability of the bacteria in a sample Thus culture techniques must be usedin combination with this cutting-edge technique to obtain a more complete snapshotof the product However as we have demonstrated here in this proof of principleanalytical sequencing is extremely promising as a first step for screening productcontents for contaminants andor mislabeling and could be potentially applied toconventional foods

Metagenomic Sequencing of Probiotics

Volume 1 Issue 2 e00057-16 msphereasmorg 9

on May 31 2020 by guest

httpmsphereasm

orgD

ownloaded from

MATERIALS AND METHODSDNA isolation Samples used for validation were grown in their respective broths as recommended bythe culture collection from which they were purchased Once grown anaerobically for 24 to 36 h theculture was pelleted (5000 g 10 min) and resuspended in 750 l of the lysis buffer included in theQiagen DNeasy kit (Qiagen Valencia CA) DNA was isolated from probiotic samples (by breaking opencapsules or direct isolation from powder forms) or the mock culture mixture with the Mo Bio PowerSoilDNA isolation mega-prep kit (Mo Bio Carlsbad CA) in accordance with the manufacturerrsquos instructionsThis kit was used because of its ability to process a large volume of sample at once which is needed toprevent clogging of the filters with capsule powder contents DNA was extracted in accordance with themanufacturerrsquos protocol DNA was then precipitated with 85 M NaCl in cold 100 ethanol and spun(10000 g) for 30 min before the supernatant was discarded Resuspended DNA (in 200 l of water) wasused for microarray analysis or sequencing Three capsules were used in each preparation to obtainapproximately 20 g of DNA

Sequencing DNA libraries were generated with the Nextera kit (Illumina San Diego CA) inaccordance with the manufacturerrsquos instructions Up to three samples were multiplexed per rundepending on sample complexity with Nextera index primers in accordance with the manufacturerrsquosprotocol and then sequenced with the Illumina MiSeq 300- or 500-cycle cartridge kit

Sequence analysis All reads were trimmed with DynamicTrim v113 prior to analysis (31) Metag-enomic sequencing data sets were analyzed by using an in-house-developed k-mer database Thegeneral approach to the development of the database has been described previously (32) Briefly acomprehensive collection of the genome sequences of commercialized species present in the normal gutenvironment was downloaded from the NCBI database (see Table S2 in the supplemental material) Thesequences were segmented into 25-bp (25-mer) nonoverlapping reads K-mers not found in at least 23of a set of additional genome sequences of the same species were removed along with any k-mer foundin the genomes of other species Typically we retained 40000 to 80000 25-mers per species Thecoverage of the 25-mers within the genome of each species was determined by testing each possiblenonoverlapping 100-bp in silico read from the genome against the 25-mers and tallying the number ofreads in the genome that were matched by the 25-mers

For the analysis regimen per se all 25-mers from each species database were placed into a triestructure Each read from a metagenomic sequencing run was compared to the 25-mers in the trie Aread in which the majority of the 25-mer matches belonged to one species was identified as that speciesThe tally of matches from the sequencing run for each species was normalized by dividing the count bythe number of in silico matched reads The k-mer algorithm compares trimmed metagenomic readsagainst these short specific 25-mers by using nucleotide BLAST searches in order to provide abundancesfor one or more sequenced metagenomes This reduces the analysis time compared to that required fora BLAST search of the full catalog of microbial genomes primarily because of the reduced size (25-mers)(33)

Microarray hybridization and data analysis The FDA GutProbe array is a custom Affymetrix DNAmicroarray platform that contains the entire gene content derived from 10 genera represented by 92whole genomes and 229 plasmids The DNA extracted from certain products in this study was subjectedto digestion labeling hybridization and scanning as previously described (9) The data were processedin R with the associated Affymetrix analysis package and the visualization of scatter plots was performedas described previously (9)

Selective culturing All products were plated on the following media for the growth of labeledingredients Luria-Bertani medium for the detection of any Gram-negative contaminants completesupplement mixture for yeast detection and potato dextrose agar for the detection of any mold etc Thelabeled serving size of each supplement was dissolved in 1 ml of water and then serially diluted 110 inMRS broth Three separate dilutions per medium type were plated with 100 l of the sample Colonieswere counted with a colony counter and colonies were picked and stained by the Gram staining method

Culture enumeration Quantification of the bacteria in a given sample is routinely achieved bydetermining the total number of CFU grown on solid medium from serial dilutions and expressed as thenumber of CFU per gram or milliliter of the original sample All of the plates were counted with a colonycounter (Advanced Instruments Norwood MA) The bacteriological analytical manual recommendscounting the colonies from the dilution that gives plate counts closest to 350 CFU and then estimatingthe total number of colonies (34) This number is then used as the estimated aerobic count The lowerlimit of enumeration can be based on the limit of detection (25 CFU from a countable range of 25 to350 CFU)

PCR detection and confirmation All single-colony isolates were incubated in 5 ml of MRS broth at37degC under anaerobic conditions for 48 h DNA was isolated from each sample by the DNeasy (Qiagen)extraction protocol for Gram-positive bacteria The DNA was cleaned and concentrated with theZymoclean DNA Clean amp Concentrator kit (Zymo Research Irvine CA) The PCR assay for strain identi-fication was performed with the Bio-Rad C1000 Touch Thermal Cycler (Bio-Rad Hercules CA) For eachreaction 15 l of DNA was added to 1 l of the forward primer 1 l of the reverse primer 2 l of the5 Omni Klentaq PCR kit and 55 l of PCR Enhancer Cocktail 1 (DNA Polymerase Technology St LouisMO) Strain-specific primers (see Table S3 in the supplemental material) were checked for specificity byusing eight control samples Lactobacillus rhamnosus ATCC 53103 Lactobacillus delbrueckii subspbulgaricus ATCC 11842 (ATCC Manassas VA) Lactobacillus casei subsp casei DSM 20080 Lactobacillusreuteri ATCC 55730 Lactobacillus plantarum 14431 and Lactobacillus casei Lc-10 The primers were thenused with single-colony DNA previously isolated from probiotic plates The PCR conditions were asfollows for all primers except for fusA initial denaturation at 95degC for 2 min 30 cycles of 94degC for 30 s

Patro et al

Volume 1 Issue 2 e00057-16 msphereasmorg 10

on May 31 2020 by guest

httpmsphereasm

orgD

ownloaded from

65degC for 30 s and 72degC for 30 s and a final extension at 72degC for 5 min For the fusA primer settemperature for the 30-s extension was decreased to 63degC All PCR products were analyzed by 1agarose gel electrophoresis

Limits of detection assays (i) DNA assay A DNA mock culture was prepared by mixing the DNAisolated from L acidophilus L plantarum L zeae L helveticus L brevis L casei and L reuteri Theconcentration of DNA was measured with the NanoDrop 21 (Thermo Scientific Wilmington DE) DNAwas diluted to the same concentration across the samples and then mixed in various proportions(Table 1) The mixture was then subjected to the Nextera library prep protocol and whole-genomesequencing was performed with MiSeq (Illumina)

(ii) Culture assay L acidophilus L plantarum L zeae L helveticus L casei L reuteri L rhamnosusand Streptococcus thermophilus were all grown separately in MRS and incubated for 48 h at 37degC underanaerobic conditions The cultures were all diluted with sterile 110 MRS medium and optical density at560 nm (OD560) was measured with a SpectraMax Spectrophotometer (Molecular Devices SunnyvaleCA) The cultures were diluted to achieve an OD560 of ~08 and mixed in various proportions (Table 1)The DNA was extracted from this culture mixture with the PowerSoil DNA extraction kit in accordancewith the manufacturerrsquos instructions (Mo Bio) The DNA was subjected to Illumina sequencing as detailedabove

Detection of E faecium in product G and isolation The sequenced product containing signaturesfor E faecium was plated on Enterococcus medium (Fisher Scientific Frederick MD) Agar plates wereincubated at 37degC aerobically overnight The single-colony isolates of E faecium were incubated inEnterococcus broth (Fisher Scientific Pittsburgh PA) at 37degC aerobically overnight DNA was isolated bythe Qiagen DNA extraction protocol for Gram-positive bacteria DNAs extracted from single-colonyisolates of E faecium were sequenced with Illumina MiSeq following preparation of the Nextera libraryThese sequences were compared to the reference strain by using seven multilocus sequence typeidentifiers

Level of E faecium detected in a noncontaminated lot of product G E faecium NCTC 7171 wasgrown overnight at 37degC aerobically in brain heart infusion medium Different loads of cells (E faeciumNCTC 7171) were used to spike the product (see Table S4 in the supplemental material) The DNA wasextracted with the Mo Bio PowerSoil extraction kit in accordance with the manufacturerrsquos suggestedprotocol The extracted DNA was subjected to Nextera library preparation in accordance with themanufacturerrsquos instructions Shotgun sequencing was performed with Illumina MiSeq (Illumina)

Nucleotide sequence accession numbers Sequences were uploaded to the NCBI BioProject siteunder ID no PRJNA291686 for the accession numbers assigned see Table S1 in the supplementalmaterial The sequences obtained by Illumina MiSeq from single-colony isolates of E faecium fromproduct G were submitted to the GenBank Sequence Read Archive (for the accession numbers assignedsee Table S1 in the supplemental material)

SUPPLEMENTAL MATERIALSupplemental material for this article may be found at httpdxdoiorg101128mSphere00057-16

TABLE S1 DOCX file 001 MBTABLE S2 DOCX file 01 MBTABLE S3 DOCX file 001 MBTABLE S4 DOCX file 002 MBTABLE S5 DOCX file 003 MBTABLE S6 DOCX file 002 MBTABLE S7 DOCX file 002 MB

ACKNOWLEDGMENTSWe thank Jayanthi Gangiredla and Isha R Patel for performing some additional k-merand microarray analyses during the review stage of the manuscript

Padmini Ramachandran was supported by a fellowship appointment administeredby the Oak Ridge Institute for Science Education

REFERENCES1 Sanders ME 2009 How do we really know when something called

ldquoprobioticrdquo is really a probiotic A guideline for consumers and healthcare professionals Funct Food Rev 13ndash12

2 Anonymous 2006 Report of a joint FAOWHO expert consultation ofevaluation of health and nutritional properties of probiotics in foodincluding powder milk with live lactic acid bacteria Cordoba Argentina1ndash 4 October 2001 Food and Agriculture Organization of the UnitedNations and World Health Organization Rome Italy ftpftpfaoorgdocrepfao009a0512ea0512e00pdf

3 Hill C Guarner F Reid G Gibson GR Merenstein DJ Pot B Morelli

L Canani RB Flint HJ Salminen S Calder PC Sanders ME 2014Expert consensus document the International Scientific Association forProbiotics and Prebiotics consensus statement on the scope and appro-priate use of the term probiotic Nat Rev Gastroenterol Hepatol 11506 ndash514 httpdxdoiorg101038nrgastro201466

4 Papadimitriou K Zoumpopoulou G Foligneacute B Alexandraki VKazou M Pot B Tsakalidou E 2015 Discovering probioticmicroorganisms in vitro in vivo genetic and omics approaches FrontMicrobiol 658 httpdxdoiorg103389fmicb201500058

5 Sanders ME Huis inrsquot Veld J 1999 Bringing a probiotic-containing

Metagenomic Sequencing of Probiotics

Volume 1 Issue 2 e00057-16 msphereasmorg 11

on May 31 2020 by guest

httpmsphereasm

orgD

ownloaded from

functional food to the market microbiological product regulatory andlabeling issues p 293ndash315 In Konings WN Kuipers OP Huis in rsquot Veld JHJ(ed) Lactic acid bacteria genetics metabolism and applications pro-ceedings of the Sixth Symposium on Lactic Acid Bacteria GeneticsMetabolism and Applications 19 ndash23 September 1999 Veldhoven TheNetherlands Springer Dordrecht The Netherlands

6 Research GV 2014 Probiotics market worth to reach $5234 billiongrowth at CAGR of 76 till 2020 new report by Grand View ResearchInc Grand View Research Inc San Francisco CA

7 Dethlefsen L Huse S Sogin M Relman D 2008 The pervasive effectsof an antibiotic on the human gut microbiota as revealed by 16S rRNAsequencing PLoS Biol 6e280 httpdx doi org101371journalpbio0060280

8 de Moreno de LeBlanc A LeBlanc JG 2014 Effect of probiotic admin-istration on the intestinal microbiota current knowledge and potentialapplications World J Gastroenterol 2016518 ndash16528 httpdxdoiorg103748wjgv20i4416518

9 Patro JN Ramachandran P Lewis JL Mammel MK Barnaba TPfeiler EA Elkins CA 2015 Development and utility of the FDA ldquoGut-Proberdquo DNA microarray for identification genotyping and metagenomicanalysis of commercially available probiotics J Appl Microbiol 1181478 ndash1488 httpdxdoiorg101111jam12795

10 Floch MH 2014 Recommendations for probiotic use in humansmdasha2014 update Pharmaceuticals 7999 ndash1007 httpdxdoiorg103390ph7100999

11 Hamilton-Miller JMT Shah S 2002 Deficiencies in microbiologicalquality and labeling of probiotic supplements Int J Food Microbiol72175ndash176 httpdxdoiorg101016S0168-1605(01)00703-6

12 Weese JS Martin H 2011 Assessment of commercial probiotic bacte-rial contents and label accuracy Can Vet J 5243ndash 46

13 Drago L Rodighiero V Celeste T Rovetto L De Vecchi E 2010Microbiological evaluation of commercial probiotic products available inthe USA in 2009 J Chemother 22373ndash377 httpdxdoiorg101179joc2010226373

14 Saxelin M 2008 Probiotic formulations and applications the currentprobiotics market and changes in the marketplace a European perspec-tive Clin Infect Dis 46(Suppl 2)S76 ndashS79 httpdxdoiorg101086523337

15 Huys G Vancanneyt M DrsquoHaene K Vankerckhoven V Goossens HSwings J 2006 Accuracy of species identity of commercial bacterialcultures intended for probiotic or nutritional use Res Microbiol 157803ndash 810 httpdxdoiorg101016jresmic200606006

16 Drisko J Bischoff B Giles C Adelson M Rao RV Mccallum R 2005Evaluation of five probiotic products for label claims by DNA extractionand polymerase chain reaction analysis Dig Dis Sci 501113ndash1117httpdxdoiorg101007s10620-005-2931-z

17 Szajewska H Fordymacka A Bardowski J Goacuterecki RK MrukowiczJZ Banaszkiewicz A 2004 Microbiological and genetic analysis ofprobiotic products licensed for medicinal purposes Med Sci Monit10mdashBR346 ndashBR350

18 Bonetta L 2010 Whole-genome sequencing breaks the cost barrier Cell141917ndash919 httpdxdoiorg101016jcell201005034

19 Hayden EC 15 January 2014 Is the $1000 genome for real Nat NewsExplainer httpdxdoiorg101038nature201414530

20 Kumar H Salminen S Verhagen H Rowland I Heimbach J BantildearesS Young T Nomoto K Lalonde M 2015 Novel probiotics andprebiotics road to the market Curr Opin Biotechnol 3299 ndash103 httpdxdoiorg101016jcopbio201411021

21 Cremonesi P Chessa S Castiglioni B 2012 Genome sequence andanalysis of Lactobacillus helveticus Front Microbiol 3435ndash 455 httpdxdoiorg103389fmicb201200435

22 Dobson CM Chaban B Deneer H Ziola B 2004 Lactobacillus caseiLactobacillus rhamnosus and Lactobacillus zeae isolates identified bysequence signature and immunoblot phenotype Can J Microbiol 50482ndash 488 httpdxdoiorg101139w04-044

23 Piwat S Teanpaisan R 2013 16S rRNA PCR-denaturing gradient gelelectrophoresis of oral Lactobacillus casei group and their phenotypicappearances ISRN Microbiol 2013342082 httpdxdoiorg1011552013342082

24 Mannu L Paba A Daga E Comunian R Zanetti S Dupregrave I Sechi LA2003 Comparison of the incidence of virulence determinants and anti-biotic resistance between Enterococcus faecium strains of dairy animaland clinical origin Int J Food Microbiol 88291ndash304 httpdxdoiorg101016S0168-1605(03)00191-0

25 Hoffmann DE Fraser CM Palumbo FB Ravel J Rothenberg KRowthorn V Schwartz J 2013 Probiotics finding the right regula-tory balance Science 342314 ndash315 httpdxdoiorg101126science1244656

26 Palmer C Bik EM DiGiulio DB Relman DA Brown PO 2007 Devel-opment of the human infant intestinal microbiota PLoS Biol 5e177httpdxdoiorg101371journalpbio0050177

27 DeAngelis KM Wu CH Beller HR Brodie EL Chakraborty RDeSantis TZ Fortney JL Hazen TC Osman SR Singer ME TomLM Andersen GL 2011 PCR amplification-independent methods fordetection of microbial communities by high-density microarray Phy-loChip Appl Environ Microbiol 776313ndash 6322 httpdxdoiorg101128AEM05262-11

28 Hamady M Knight R 2009 Microbial community profiling for humanmicrobiome projects tools techniques and challenges Genome Res191141ndash1152 httpdxdoiorg101101gr085464108

29 Sun Z Zhang W Guo C Yang X Liu W Wu Y Song Y Kwok LY CuiY Menghe B Yang R Hu L Zhang H 2015 Comparative genomicanalysis of 45 type strains of the genus Bifidobacterium a snapshot of itsgenetic diversity and evolution PLoS One 10e0117912 httpdxdoiorg101371journalpone0117912

30 Dicks LM Du Plessis EM Dellaglio F Lauer E 1996 Reclassification ofLactobacillus casei subsp ATCC 393 and Lactobacillus rhamnosus ATCC15820 as Lactobacillus zeae nom rev designation of ATCC 334 as theneotype of L casei subsp casei and rejection of the name Lactobacillusparacasei Int J Syst Bacteriol 46337ndash340 httpdxdoiorg10109900207713-46-1-337

31 Cox MP Peterson DA Biggs PJ 2010 SolexaQA at-a-glance qualityassessment of Illumina second-generation sequencing data BMC Bioin-formatics 11485 httpdxdoiorg1011861471-2105-11-485

32 Leonard SR Mammel MK Lacher DW Elkins CA 2015 Application ofmetagenomic sequencing to food safety detection of Shiga toxin-producing Escherichia coli on fresh bagged spinach Appl Environ Mi-crobiol 818183ndash 8191 httpdxdoiorg101128AEM02601-15

33 Altschul SF Gish W Miller W Myers EW Lipman DJ 1990 Basic localalignment search tool J Mol Biol 215403ndash 410 httpdxdoiorg101016S0022-2836(05)80360-2

34 Anonymous 1998 Bacteriological analytical manual US FDA SilverSpring MD httpwwwfdagovFoodFoodScienceResearchLaboratoryMethodsucm2006949htm

Patro et al

Volume 1 Issue 2 e00057-16 msphereasmorg 12

on May 31 2020 by guest

httpmsphereasm

orgD

ownloaded from